Execute Models Anywhere

Whether performing a single decision or millions, the decision engines feature parallel code execution and a built-in low-latency web stack that maximise throughput and high scalability.

• Execution speed through code generation and execution without the need for remote model calls
• Horizontal and vertical scalability and high availability
• One-click deployments of models and roll-backs without downtime
• Simultaneous hosting of multiple model versions (dynamic routing, e.g. for champion/challenger)
• Stateless architecture allows scaling across multiple engines with elastic load balancing (e.g. in Kubernetes)

ACTICO allows for seamless decision execution in near real time. All decisions can be called easily from your application as web services.

• REST, JSON/XML, SOAP and OpenAPI interfaces for centralized decision services
• Low latency web stack for thousands of decisions per second
• Multiple simultaneous model versions, including API versioning

The Java based execution engine accelerates decisioning by enabling customers to run natively in their environment (in-process automated decisioning within Java applications).

• Lightweight execution engine for models
• Integration via native Java API allows direct embedding of decisions into Java applications
• Dynamic update of models from a central repository, without the need to rebuild or restart the application using it
• Optimised memory footprint to simultaneously handle many large models

ACTICO Decision Management Platform is not a “black box” when it comes to decisions. You can reproduce every decision made, even years later.

• Visualisation of the full execution path within the graphical models
• Dependency graph of all contributing models and versions
• Explanations of individual machine learning predictions
• Trace data can also be published to external systems (e.g. Kafka)